Without packet loss, delay or jitter in an IP network, Voice over IP (VoIP) calls typically do not suffer from voice degradation problems. The quality of VoIP however, is typically degraded when any delay occurs or if packets are misrouted or lost.
In real-time applications such as, for example, VoIP, a delayed packet is also considered to be a “lost” packet. At the receiver side, the effect of a “lost” packet on the degradation of voice quality can be reduced by an effective packet-loss concealment (PLC) algorithm. For example, one conventional method includes either muting the output speech for the time interval of the missing speech frame or replacing the missing samples in the missing packet with the previous speech frame. Conventional methods, however, can result in distortions which disturb the listener even at low packet loss rates.
Other conventional methods are generally computationally intense and complex. For example, one method generates approximations of the missing speech samples. Other conventional methods use PLC algorithms that are vocoder dependent and are most effective when the missing speech packet is encoded using a particular vocoder. Code Excited Linear Prediction (CELP) based vocoders, for example, have speech parameters such as Linear Prediction coefficients and pitch periods encoded in them. PLC algorithms for these vocoders are typically based on repeating these speech parameters for the missing speech frame.
Still other conventional methods include using high-quality PLC algorithms based on pitch-detection. Some of these methods are used when complying with standards from the International Telecommunication Union (ITU) such as, for example, the G.711 standard, and are can be used with audio compression techniques. In particular, the G.711 standard represents the logarithmic pulse-code modulation (PCM) encoding/decoding standard (as part of Appendix I).
Voice quality may be evaluated using a Mean Opinion Score (MOS) testing technique. The MOS technique employs the use of live listeners and thus is costly and requires a good deal of time to perform. Automated techniques such as, for example, Perceptual Evaluation of the Speech Quality (PESQ), assess voice quality relatively rapidly and at less expense than most MOS techniques.
Regardless of the algorithm used to predict the missing samples, the final objective of most conventional PLC techniques is to improve voice quality.